基于改进粒子群算法和元胞自动机的城市扩张模拟——以南京为例

为合理利用多智能体算法解决城市扩张动态模拟问题,基于地理学理论和社会学规律对粒子群算法进行有针对性的改进,提出分段式粒子群算法(SPSO),并结合元胞自动机模拟复杂时空过程的能力,构建出适用于城市扩张模拟的地理元胞自动机SPSO-CA。在SPSO-CA中我们利用多时像的土地利用数据、交通路网数据和地形数据,挖掘出1995~2000年南京城市扩张的土地转换规则。再由此规则实现1995~2008年的南京市城市扩张过程的动态模拟。最后对比SPSO-CA、PSOCA及NULL模型结果得:SPSO-CA总精度86.3%,Kappa系数为0.792,Moran’s I为0.078,PSO-CA总精度83.6%,Kappa系数为0.755,Moran’s I为0.054,NULL模型总精度81.9%,Kappa系数为0.741,真实的Moran’s I为0.072。这表明无论是总精度还是空间一致性,SPSO-CA都优于PSO-CA和NULL模型,即用SPSO-CA模拟城市扩张是可行的。

URBAN EXPANSION SIMULATION USING MODIFIED PARTICLE SWARM OPTIMIZATION ALGORITHM AND CELLULAR AUTOMATA: A CASE STUDY OF NANJING CITY

For scientific use of multi-agent algorithm to model dynamic urban growth,Subsection Particle Swarm Optimization (SPSO),an improved algorithm has been proposed in this paper.The improvement is based on the general rule in geography and sociology.Cellular Automata is also combined to simulate complex spatial-temporal processes.An new Geographic Cellular Automata (SPSO-CA) is constructed to achieve the dynamic simulation of urban growth.Deriving transition rules is key to the Geographic Cellular Automata.We therefore discover first the transition rules for SPSO-CA based on 1995-2000 land use data,traffic network data and terrain data.And then,dynamic simulation of urban expansion process of Nanjing City from 1995 to 2008 is made according to this rule.Lastly,in order to test the effectiveness of the improved algorithm,we compared SPSO-CA,PSO-CA and NULL model,the following results were obtained.The overall accuracy of SPSO-CA is 86.3％,with Kappa coefficient of 0.792,Moran's I of 0.078;the overall accuracy of PSO-CA is 83.6％,with Kappa coefficient of 0.755,actual Moran's I of 0.054;the overall accuracy of NULL model is 81.9％,with Kappa coefficient of 0.741,with actual Moran's I of 0.072.These results demonstrate that SPSO-CA is better than PSO-CA and NULL model and the improvement of Subsection Particle Swarm Optimization is available.